Study On Hot-air Drying Prediction Model Of Ginger Based On BP Neural Network And Support Vector Machine

Ginger is hot air drying on the fresh ginger for the preservation of one of the main instruments, using experimental study on drying characteristics of the ginger, ginger drying process and changes of water content in a forecast that actual production is of great importance.Study this paper mainly discusses two aspects, first through the ginger drying single-factor experimental study on drying characteristics of ginger, then using BP neural networks and SVM methods to changes in moisture content in drying process of fresh ginger for prediction. By Ginger drying characteristics, analysis of the ginger drying process and hot air, hot air speed and slices the thickness of these three factors effect on the drying properties of ginger. After variance analysis of orthogonal experiment, air temperature, air speed and slice thickness on the drying rates are having an effect of ginger, from large to small are in turn affected slice thickness, the hot air temperature and air speed. The ginger drying process using BP neural network prediction model of water content, getting the model to be fit for the0.99981, the training samples of mean square error for0.00013728, test samples obtained a maximum absolute error of22.58%, mean absolute percentage error is20.23%. Building the ginger drying process using support vector machines prediction model of water content, getting the model to be fit for the0.99916, the training samples of mean square error for0.00006091, test samples obtained a maximum absolute error of1.62%, mean absolute percentage error is1.12%. Results showed that using support vector machines prediction model of BP neural network to predict building model forecasting errors are smaller and prediction accuracy is higher.